Database Reference
In-Depth Information
By having dedicated servers for each application, we ensure those applications have the
resources they need (CPU, memory, and disk) when they need them. It also means that a
lot of very expensive hardware is sitting idle. According to the VMware literature, 8%-
12% average utilization is typical. We cannot speak to that number directly. What we
can speak to is that when we undertake one of the many database-consolidation
engagements we do each year, our experience has taught us that 60% database-
consolidation ratios would be a very conservative number. Our clients typically
experience better than 60% database consolidation ratios. This means we are able to
place all their current databases in half the footprint they consume today. There is a lot
of wasted spare capacity out there, which tells me if we were to look across a large
organization, then 12% overall utilization would not be that far off, especially if you
take weekends into consideration. I cannot think of a bigger workhorse than a server
managing a database—and look at all the wasted capacity out there. Therefore, the 12%
number rings true.
Having a one-to-one relationship between applications and hardware is a very
expensive proposition, yet this was the world before virtualization.
This was driven in part by the inability of the operating system's resource management
capabilities to prevent different applications from impacting each other. This was
further complicated by incompatible DLLs and libraries between different applications
running on the same operating system. This was a world we lived in because we did not
have options in the past. It's a world a lot of companies still live in. It is a very
expensive world to live in. Cost and business agility are some of the many drivers why
companies are virtualizing more and more of their infrastructure.
One to Many: The Virtualized World
In the new computing paradigm of virtualization, it is a one-to-many relationship, where
a single physical server will now have many applications residing on it, improving
overall utilization rates but still offering the same level of service.
If you refer back to
Figure 3.1
, you'll see that it illustrates the one-to-one world before
virtualization. Before virtualization, you would have three separate physical servers
handling all the work. If those servers were only busy 50% the time, which is being
generous, then the other half of the time the available resources such as CPU and
memory went to waste. The other physical servers have no way of taking advantage of
that spare capacity no matter how badly they need it.
In
Figure 3.2
,
on the other hand, we illustrate a virtualized infrastructure where every
one of those physical servers is now represented by different virtual machines on a
single host. When there is spare capacity, the other VMs are able to take advantage of
those resources if they are in need. Another advantage of virtualization is that those
three physical servers could be the right size on a physical host containing the capacity
